I am hopeful that you have studied and explored all the topics of machine learning and at the same time practiced assignments also.
In today's tutorial we will be discussing ten questions that may be asked in interviews. You have to listen to these questions carefully and understand and after that whatever topics we have learnt in the learning course you have to prepare somewhat in this form to answer it in interviews. So let us now begin with the first question.
Our first question is what is machine learning? meaning what machine learning is?. So to answer this question you should somewhat tell in this way.
Machine learning is an ability in machine to learn from data and to identify patterns without human intervention. meaning that machine learning is an ability that is imparted to the machine or given to the machine that by observing data or by reading data it can learn , identify patterns from it and in this whole process human intervention meaning not much requirement or intervention by humans should not be there and this machine should be able to follow on its own. So this is called machine learning.
This happens in exactly the similar-way in which we humans see a new thing or while having an experience store that experience in our mind. Not only this much, from that experience we derive some inferences also. Our method of forming an attitude approaches and takes actions that are also affected by that particular experience. So similarly machines also data, meaning the data is an experience for the machine , so machines also learn and experience from this data and they also give an output , and their net output is based on the last experience and this is called machine learning.
You should always give an example also along with your answer so that it is very easy to understand and make people understand things easily. So now if we talk about automatic vacuum cleaners , then if an automatic vacuum cleaner by seeing dust can identify that this is dust then it means that it is a machine learning application. So in this way by giving one or two examples you can conclude your answer. Now let us move to the next question.
Our next question is what are different types of machine learning algorithms. Which are the different types of algorithms in machine learning or what are they? There are three major algorithms of Machine learning.
The first type is supervised, the second type is unsupervised and the third type is reinforcement learning. Supervised learning is also called task driven because that data that we have is labeled data in which we have the X meaning input features and Y meaning output features. It has examples like k nearest neighbor .
As the name suggests you will be able to visualize that in this the new data input that has been observed, which are the nearest inputs to it and when we are aware of the nearest neighbors then we see their class , that what is the class of those neighbors and the one which has the maximum class and has the match we pick that one , so this is an example of K nearest neighbor. In the same manner in support vector machines we have a classifier also and if you move further so in neural networks also if you are doing feed forward along with back propagation there also it becomes a supervised approach because in back propagation the previous output is known , so this was supervised.
The second technique is unsupervised technique. Unsupervised learning is data driven meaning that data drives it, In this we do not have input and output of data in the form of X and Y , meaning that it is unlabeled data, there are some features that we can call X1, X2 X3 and upon that we run the algorithm to learn the patterns, this is how unsupervised learning works.
For this we have the K means algorithm as an example. Apart from K means we can graphically form clusters or by forming clusters hierarchically also we can implement its algorithms.
Now if we talk about reinforcement learning , now reinforcement learning is learning to react to an environment means here we are not talking of any labeled or unlabeled data .
It is taking only fresh inputs , although there are rules already set as to what action it will take , what the strategy will be like, so these rules are already set in reinforcement learning but after taking that action whether you are get a reward or you get a punishment or penalty , so n the basis of these rewards and penalties this algorithm learns how to adopt a better way and here Q learning is a better example.
So this was our question number two , so now let us move further to our next question which is question number three, what is supervised learning.
What is supervised learning? Supervised learning is a machine learning algorithm of inferring a function from labeled training data. So supervised, in its name only the answer is hidden, supervised means that you are able to supervise something.
You have some input and what was its output that you can supervise means you are able to see what was the output and doing this you saw one record, two records and like this you saw hundreds and hundreds of records. Now after seeing what the mind does , the mind will take out inferences from every record and will create a belief.
So this is how supervised learning algorithms do. It tries to create a function that matches the input and output. When you map input and output, so by mapping a function is created , now if you have only one input, in future you receive only one input so the mapping by that time has become so strong that the output that it gives for this matches exactly with the real output.
So this is how these supervised learning algorithms carry out predictions. Examples of these are support vector machines, in regression we have support vector regressors, Naive Bayes algorithms, Decision trees classifiers and regressors, K nearest neighbor algorithms , so these are some of its examples. So let us now move towards our fourth question.
And our fourth question is unsupervised learning. Unsupervised learning is also a type of machine learning algorithm used to find patterns on the set of data given.
o the primary target of unsupervised learning is that it identifies patterns, you already have data but you do not have any label with it , with labels you get help in knowing patterns and you can record it as a label and write it , but here there are no labels, so unsupervised learning algorithms performs on data without labels and the similarity in features between data , it creates clusters or groups on basis of that similarity and by creating groups it basically decides as to what action is to be taken ahead.
So in this there can be clustering, anomaly detection and neural networks also follow unsupervised learning and with this we will now move to the fifth question and the fifth question is what are support vectors in SVM.
So, in SVM what are the support vectors? So support vector machines work on support vectors , and how does it work? The support vector machine first chooses support vectors and for its hyperplanes it creates margins , so the data points used to create margins are called support vectors.
So support vector machine is an algorithm that tries to fit a line or plane or hyperplane between different classes that maximizes the distance from the line to the point of classes., so what is the meaning of point of classes , the data points that are there, so SVM tries to make a line, a plane or a multi dimensional hyperplane , so what does that plane and hyperplane do, it basically differentiates between one class from another class, but from that hyperplane that class starts at what distance or the data points start from what distance, and the most nearest data point is called support vector.
So let us now move to our next question which is question number six and this question is what are different kernels in SVM. So in SVM algorithms , what are kernels?
So kernels are functions in algorithms which are developed while creating matching functions or they help in this and these are of six types. Linear kernel, polynomial kernel, radial basis kernel, sigmoid kernel are four out of them.
Now these functions, linear kernel is a function , polynomial is a function , so these functions we use according to the type of data we have , so depending on type of data we decide that which kernel to use like for linear kernel our data should also be in linear fashion otherwise use a polynomial kernel , so in this way we have seen what are kernels and what are its types , that we have seen and now well jump to question number seven.
Question number seven is what is Bias in machine learning?. Bias in the data tells us that there is inconsistency in the data.
The inconsistency may occur for several reasons that are not mutually exclusive.
So here through an example let us understand, a tech giant like amazon to speed the hiring process they build one engine that they are going to give hundred resumes from which it will spit out the top five and hire those.
Now understand that this particular software that amazon made that you can give it soft copies of resume and after giving soft copies the output that will be received will be only five six resumes which will be the best but at the time of using it a glitch was encountered , now the glitch was that it was giving output such that it will give one output for female and four for males and at time all outputs were for males , so this is a bias, it's a bias, so in this way bias is introduced in data sets due to some process and that is called Bias in machine learning which affects the output also. So let us now move towards our eighth question.
Eighth question is to explain the difference between classification and regression. So while giving the answer for this you can start with classification and regression they both come under supervised learning but, but they have differences , the difference is classification algorithm works on data that is categorical and regression works on data that is continuous so first and basic difference between classification and regression is this only.
Now when the data is categorical then classification algorithm divides data into classes and these classes are known already class wise that whether there are four classes or five classes , so data belongs to those classes only but in case of regression what happens is that in this no fixed value is taken as target and there you apply regression and the values that is there are brought near to values already given through averaging. So in this way classification and regression work on two different problems.
As an example if I want to know that in a match who will be first, who will be second, who will be third , so for this I wrote a predictive algorithm and I calculated and saw that who will be first, who will be second and who will be third and my algorithm answered it, so what happened with this, what happened is that the one who is first is from first class, the second is from second class and the third means third class , so these are classes , so in this way classification works , but if I go to buy a flat , so at the time of buying the flat it's current price was seventy lac rupees , tomorrow it grew to seventy three and day after it may become seventy five also , so in this way what happens is that in our regression problem we run an algorithm to find continuous data, how the predictions for continuous data can be done , so now we will move towards the next question which is the ninth question.
And this question is explain the terms artificial intelligence, machine learning and deep learning. Explain what are artificial intelligence, machine learning and deep learning. Although these terms are correlated with each other, even then if you want to distinctly identify them , then artificial intelligence is the domain of producing intelligent machines.
Artificial intelligence is that part that focuses on producing artificial machines. Machine learning refers to systems that can assimilate from experience , machine learning is that in which a machine program on being fed from computer data starts learning and after learning it starts acting without any human intervention. Deep learning refers to or states to systems that learn from experience on large data sets .
Deep learning, it is inspired from human brain, this works on learning from experience because here the complete data is not given initially and it has almost zero data and taking actions it learns if it is running in reinforcement approach but if we are running deep learning with back propagation then it will have data also and it will become supervised , otherwise we can run it in unsupervised way also , so this is artificial intelligence and subsets of it are machine learning and it's subset is deep learning..
So let us now focus on the last question of this video and this question is what is the role of encoders in neural networks. In neural networks, what is the role of encoders? Encoders convert categorical data into numerical format.Encoders are very important.
They convert the categorical data into numerical format.For example if I have red, red, yellow and blue and I encode it and let us assume that red becomes one , the next becomes two and the third becomes three then I will have the data one ,one, two ,three , in this manner, so here we have converted one string type data, a categorical type data and what we did with it, what we did, we converted into numerical format , so this is called encoders , so encoders are in preparation layer and when you want to implement machine learning algorithms , so we use encoders at the time of data preprocessing , to make data healthy, for processing purpose. So these were the ten questions of this video. I wish you best of luck for your next job hunt or job change and keep revising concepts in this manner.
If you have any queries or comments, click the discussion button below the video and post there. This way, you will be able to connect to fellow learners and discuss the course. Also, Our Team will try to solve your query.
Ruturaj Nivas Patil
Very well explained in entire course. Great course for everyone as it takes from scratch to advance level.